In fight against hate speech, FB's new AI tool can detect offensive memes

Don't want to miss the best from Business Standard?

In a bid to curtail offensive memes that promote hate speech, Facebook is building a novel artificial intelligence (AI) system that uses machine learning to identify text in images and videos as well as transcribe it.

While tools to transcribe text are nothing new, the company faces different challenges because of the size, the sheer volume of photos shared each day on Facebook and Instagram, and the number of languages supported on its global platform.

With the novel AI system called Rosetta, Facebook can process more than a billion public images and Instagram images and video frames (in a wide variety of languages) per day through the system efficiently.

The extracted text is then used by downstream classifiers to immediately act upon the policy-violating content or by-product applications like photo search.

"Understanding text in images along with the context in which it appears also helps our systems to proactively identify inappropriate or harmful content and keep our community safe," the social media giant said in a blog post late on Tuesday.

"Text extracted from images is being used to improve the relevance and quality of photo search, automatically identifying content that violates our hate-speech policy on the platform in various languages, and improve the accuracy of classification of photos in News Feed to surface more personalised content," the networking giant noted.

Further, Facebook is also continuing to invest in extending the text recognition model for the wide number of languages used on its global platform.